PSCBO: A provably secure consensus-based blockchain Oracle

Abstract

Blockchain technology, successful in cryptocurrency, holds transformative potential for various industries. In decentralized applications driven by blockchain, acquiring non-digitally verifiable data is crucial, particularly in smart contracts. To facilitate the transfer of off-chain data from websites to the blockchain, several studies have proposed various centralized and decentralized blockchain Oracles. However, centralized Oracles reintroduce central risks, such as the Single Point of Failure, while decentralized Oracles consistently rely on a voting mechanism, which incurs additional vulnerabilities and substantial costs. In this work, we uncover a misdirection attack in decentralized Oracles, resulting in a diminished security region in blockchain consensus. This manifests as prolonged settle times and reduced upper bounds for tolerance of adversarial consensus chip. By formalizing the voting process as a discrete-time Markov chain (DTMC), we further demonstrate that the maximum stake of the adversary for launching a misdirection attack is less than 50%. To counteract this threat, we introduce a novel voting model that relies on chain length rather than block data for voting. Formally, we propose the definition of our model named $VoteFork$. Following this, we present the specific consensus-based blockchain Oracle protocol, PSCBO. This protocol involves blockchain clients collecting and providing Oracle data, with an epoch-based voting mechanism to verify each set of Oracle information. We also analyze the security of the PSCBO backbone $VoteFork$, demonstrating its security region akin to a typical blockchain. Finally, we implement a proof-of-concept system to validate the security region and further underscore the practicality. Both security analysis and experiments affirm that PSCBO exhibits a higher adversarial fault tolerance and shorter waiting time to confirm transactions.